IT, Cloud & DevOps Blog

visa vulnerability agentic harness ai patching a upright shield pointed base

Visa Ships a Security AI That Patches Your Code Before Any Human Reviews It

Visa has open-sourced the Visa Vulnerability Agentic Harness, the tooling it used to hunt for bugs across its own payment network, and its documentation contains a line worth pausing on: “A plain scan edits your code.” In its default configuration the harness runs discovery, remediation and validation end to end, writing candidate fixes straight into source files before any person has looked at the finding or the patch. Nothing reaches production without a human, but nobody approves the patch before it is written. This breakdown covers the four-phase, eleven-stage pipeline, exactly where the “no human review” claim holds and where it collapses, what Visa itself says about oversight, and the independent evidence that only 26% of AI-generated security patches fix the flaw without changing what the software does.

Read more
project ot meta ai restructuring layoffs a upright chess rook piece

Meta Reportedly Abandoned an AI-Focused Restructuring Plan That Would Have Laid Off Thousands

Project OT — short for Organization Transformation — was Meta’s plan to make the company “AI native” by handing much of its routine work to software agents, leaving small “talent-dense” pods of humans to supervise them. A Reuters special report published on 26 August 2026 says the plan was drawn up at Mark Zuckerberg’s Hawaii compound in January, modelled team cuts of up to 60 per cent, and was abandoned in part on the evening of 19 May, hours before the first wave of notices went out. This breakdown covers what Project OT actually was, the pod and village structure it introduced, the internal metrics that undermined it, the keystroke-tracking software that triggered an employee revolt, the 8,000 layoffs and 7,000 redeployments that did happen, and what any organisation writing an AI workforce plan should take from it.

Read more
nscale anthropic 45 billion vera rubin a lattice electricity pylon

Anthropic to Pay Nscale $45 Billion for Power and Nvidia Vera Rubin Chips

Nscale has agreed to sell Anthropic roughly $45 billion of AI computing capacity over six years, according to Bloomberg and CNBC reporting on 26 August 2026. The capacity amounts to about 460 megawatts at a West Virginia campus, fitted out with Nvidia Vera Rubin systems arriving from late 2027. This breakdown covers the reported terms, what 460 megawatts actually means, who Nscale is and how a $45 billion contract reshapes its September listing plans, the Monarch Compute Campus and its off-grid gas microgrid, the Vera Rubin generation-on-generation numbers, where the deal sits alongside Anthropic’s Google, Broadcom, Amazon, Microsoft and SpaceX commitments, the price per megawatt-year, the risks nobody is pricing, and what any of it means for organisations that buy AI rather than build it.

Read more
querystory ai data analysis trust gap a magnifying glass upright

QueryStory Wants You to Believe What AI Is Telling You: Inside Its $6M Bet on Verified Answers

QueryStory came out of stealth on 26 August 2026 with a $6 million seed round from Brightmind Ventures and New York Life Ventures at a $60 million valuation, and an unfashionable pitch: the problem with enterprise AI analytics is not speed, it is that nobody can tell whether the answer is true. This breakdown covers CEO Shapor Naghibzadeh’s route from Google’s Operation Aurora war room through six years of security tooling to co-founding Chronicle in Google X Labs, the founding team alongside CTO Stanley Yang and CPO David Glusic, the four mechanisms the platform is built on — SQL that surfaces automatically, an explicit confidence indicator, human review recorded in the platform, and narrative assembly that refreshes as the data moves — the argument against consumption-priced frontier-lab tools, TechCrunch’s hands-on test that produced in a few hours a space-activity visualisation that once took several weeks with a developer, the arithmetic of a $6M round at a $60M valuation, a due-diligence table for anyone evaluating AI analytics, and the four questions the launch leaves genuinely unresolved.

Read more
Qwen3.8-Flash - qwen3 8 flash next 125b moe model a honeycomb block seven cells

Alibaba Releases Qwen3.8-Flash: A Multimodal 125B MoE Model That Previews Qwen4

Alibaba open-weighted Qwen3.8-Flash-Next on 26 August 2026: a multimodal mixture-of-experts model with 125 billion parameters, a separate 51-billion-parameter N-gram embedding table, and just 6 billion parameters activated per token. This breakdown covers the four rebuilt subsystems — Gated DeltaNet paired with Qwen Sparse Attention at block granularity, a Gated Residual stream widened to four gated branches, the N-gram table that offloads to host RAM, and the Muon plus AdamW training recipe with batch-size warmup removed — alongside the 48-layer stack of 512 experts that fires eleven per token, the published benchmark table showing 62.5 on SWE-bench Pro against 53.4 for Claude Opus 4.6 and 84.5 on AndroidWorld against 62.0, the single loss on Humanity’s Last Exam at 35.9 against 40.0, the unverifiable one-ninth training cost claim, the 262,144-token native context extended to a million with YaRN, hosted pricing of $0.16 and $0.47 per million tokens against $2.00 and $6.00 for Qwen3.8-Max, the real hardware bill from a 172.78 GiB FP8 checkpoint down to a 111 GB four-bit GGUF, the qwen-community-1.0 licence that is not Apache 2.0, and a buyer’s checklist for treating a preview checkpoint as a production dependency.

Read more
CHAT